Genopolitics and the Science of Genetics

Home , Genopolitics

American Political Science Review Page 1 of 14 May 2013 doi:10.1017/S0003055413000099 Genopolitics and the Science of Genetics EVAN CHARNEY Duke University WILLIAM ENGLISH Harvard University nanearlierarticlewechallengedthefindingsofFowlerandDawes(FD)thattwogenespredictvoter turnout as part of a more general critique of “genopolitics.” FD now acknowledge that their finding I of a “significant” direct association between MAOA and voting was incorrect, but claim to have replicated their finding of an “indirect” association between 5HTT, self-reported church attendance, and self-reported voting. We show that this finding is likely driven by population stratification and omitted variable bias. We then explain why, from the standpoints of genetics, neuroscience, and evolutionary biology, genopolitics is a fundamentally misguided undertaking; we also respond to FD’s charge that some of our previous statements concerning genetics are “highly misleading,” “extremely disingenuous,” and “even incorrect.” We show that their criticisms demonstrate a lack of awareness of some basic principles in genetics and of discoveries in molecular genetics over the past 50 years.

ewouldliketothanktheeditorsofThe Amer- STATISTICS ican Political Science Review for inviting us Wto participate in this Forum by writing a re- Population Stratification sponse (or “rejoinder”) to the articles of Fowler and Dawes and of Deppe, Stoltenberg, Smith, and Hibbing. Fowler and Dawes (FD) acknowledge that their previ- We view this as a welcome and important opportu- ous, highly publicized finding that a polymorphism of nity. Although we address both articles, our emphasis the MAOA gene showed a “significant” direct associ- throughout is on the contribution of Fowler and Dawes, ation with (self-reported) voter turnout was incorrect, which itself is intended, in part, as a rejoinder to our inasmuch as they failed to replicate it using new data earlier article in this journal, “Candidate Genes and from Wave IV of the National Longitudinal Study of Political Behavior” (Charney and English 2012). Adolescent Health (Add Health) dataset. They give Our response is divided into two parts. Part I, “Statis- little consideration as to why their finding that “indi- tics,” is an empirical critique of the specific gene- viduals with a polymorphism of the MAOA gene are behavior association claimed by Fowler and Dawes significantly more likely to have voted in the 2004 presidential election” (Fowler and Dawes 2008, 579) did not and, in a much weaker version, by Deepe et al. Al- 1 though our statistical critique addresses specific claims, replicate, viewing it simply as a rare false positive. In- as did our previous article, it also points to broader stead, they focus attention on the “second gene” from problems that likely affect all gene association studies their initial assertion that “two genes predict voter in behavior genetics. Part II, “Genetics,” is a response turnout;” namely, that a polymorphism of the 5HTT gene, when interacted with (self-reported) frequency to specific charges leveled by Fowler and Dawes against 2 several of our assertions regarding genetics and the of church attendance, predicts (self-reported) voter nature of the genotype-phenotype relation in regard turnout. FD report that they confirmed this initial find- to complex behavioral traits. It is also an extended ing with Wave IV data and take this result as providing explanation as to why the search for genes that can strong support for the larger project of genopolitics. predict complex human behaviors (such as all political However, on investigation, this result is likely false as behavior) is a fundamentally misguided undertaking. well, driven by population stratification and omitted To the reader accustomed to a cursory explanation of variable bias. genetics in an initial paragraph followed by extended There are strong empirical and theoretical reasons statistical analysis, Part II will doubtless seem unfa- to conclude that the association between voting and the interaction of “long” (L) 5HTT3 with Church miliar and perhaps unwelcome. What is so much ge- 4 netics doing in a political science journal? The answer Attendance is a spurious correlation caused by is that genetics is not a subdiscipline of statistics, and to the extent that practitioners of “genopolitics” claim 1 If failure of replication is indeed the standard for determining a to advance the science of genetics, they must be held false positive, then most gene associations in behavior genetics are accountable to that very science. false positives. For more on this, see the later discussion. 2 For overreporting of church attendance see, e.g., Marler and Hadaway (1999) and Tom (1998). 3 In what follows, we use FD’s coding for the “long” version of the Evan Charney is Faculty Fellow, Duke Institute for Brain Sciences; 5HTT gene (i.e., the alleles “long-long” [LL] and “long-short” [Ls] Faculty Researcher, Duke Institute for Genome Science and Policy; are both counted as “long.” We refer to LL and Ls as (L)5HTT, and Associate Professor of the Practice of Public Policy and Polit- unless otherwise noted. ical Science, Duke University, Rubenstein 250, Durham, NC 27708 4 Subjects reported how often they attended church, synagogue, tem- ([email protected]). ple, mosque, or other religious services in the past 12 months. The William English is a Lab Fellow, Edmond J. Safra Center for categories for response were “never,” “a few times,” “several times,” Ethics, Harvard University, 124 Mount Auburn Street, Cambridge, “once a month,” “2 or 3 times a month,” “ once a week,” and “more MA 02138 ([email protected]). than once a week. ” FD (2008, 584) report that they “simplified these Both authors contributed equally to this article. responses by grouping them into three categories of attendance:

1 Genopolitics and the Science of Genetics May 2013 population stratification. Population stratification (PS) once PS is controlled for using the first 20 principal occurs when there are systematic differences in allele components (Yashin 2011). frequencies in subpopulations due to genetic ancestry That FD pay little attention to PS is puzzling given (Bouaziz, Ambrose, and Guedj, 2011). In case control that (1) the behaviors they investigate—church atten- studies, PS can lead to spurious associations if differ- dance and voting—are known to vary widely by eth- ences in allele frequency between cases and controls nic ancestry (Chatters et al. 2008; U.S. Census Bureau are due to systematic differences in genetic ancestry 2012) and (2) 5HTT allele frequencies, both “long” rather than a causal genotype-phenotype relationship. and “short,” exhibit significant inter-ethnic variation Controlling for race/ethnicity or analyzing results by (Lotrich, Pollock, and Ferrell 2003; Noskova et al. ethnic group is one small step researchers can take to 2008). When we observe different levels of 5HTT al- try to mitigate the confounding effects of PS. However, leles among self-reported voters and nonvoters who as FD are aware, simply examining race is inadequate differ in their self-reports of church attendance, the because it does not address intra-ethnic PS, something presence of PS is the first explanatory hypothesis that that has been shown to occur even in populations should come to mind. Add Health contains data that thought to be highly homogeneous, such as Icelanders can be leveraged to investigate the likely prevalence (Helgason et al. 2005; Price et al. 2009). and effects of PS. Thus we first examined whether the One of the most commonly cited examples of a (L)5HTT∗Church Attendance association replicates false association produced by PS is the link between in more ethnically homogeneous subpopulations and the dopamine receptor gene DRD2 and alcoholism. then tested for improbable associations in the Add Initial case control studies suggested a strong associa- Health data, associations that should exist only in the tion, but subsequent investigations found none when presence of PS. more effective controls for PS were imposed (Gel- The Add Health questionnaire asks respondents ernter, Goldman, and Risch 1993). In retrospect, it to identify their “family origins”/ “family ancestries” is clear why this initial result was vulnerable to con- from a list of countries, groups, and geographic ar- founding due to PS: DRD2 alleles vary widely by eas. Respondents are allowed to identify up to four ethnic ancestry, and ethnic differences in alcoholism choices, but if more than one is chosen they are asked rates are pronounced (Thomas and Witte 2002). Early to indicate which best describes their family origins. follow-up studies of DRD2 and alcoholism used two There are significant limitations to this data: One- strategies to mitigate the confounding effects of PS: third of respondents (33.3%) indicate “America” as family-based designs and investigations restricted to the country that best describes their family origins, ethnically homogeneous subpopulations. As FD note, and approximately 7% indicate “Africa.” Thus, more family-based designs are expensive, and it is often than 40% of respondents indicate ancestral origins difficult to gather sufficient data for them to be well that from the perspective of population genetics are powered (Add Health does not contain such data). particularly uninformative. However, we were able to Sampling ethnically homogeneous populations can be examine the (L)5HTT∗Church Attendance association more feasible, but they are hard to isolate, particularly within ethnically distinct groups among the remaining in a country such as the United States, and, as noted, PS 60%. can still occur within populations that appear racially Table 1 shows the logit coefficients for homogeneous. (L)5HTT∗Church Attendance associations within Genetic epidemiologists have long noted that “even each of the 10 largest self-identified ethnic groups, small stratification can have considerable conse- using FD’s mixed-effects model with standard controls quences for large samples” (Devlin, Bacanu, and for age, sex, and race. Although these samples are Roeder 2004, 1129). One of the benefits of genome- small and thus underpowered, the extraordinarily wide association studies (as contrasted to candidate heterogeneous effects across ethnic groups suggest the gene studies) is that genome-wide data can be lever- likelihood of PS. More than half of the results show aged to help control for PS directly (Bouaziz, Ambrose, an association that trends in the opposite direction and Guedj 2011). Currently, the most effective strategy from FD’s finding, including the largest ancestry group for dealing with PS involves using a large range of (Germany, N 1,289). Moreover, the only result that single nucleotide polymorphisms (SNPs) contained in is significant is= negative (Philippines: .94, p .03). genome-wide data and employing techniques such as Although the self-identified primary country− of= family genomic control, structured association, and principal origin is a crude measure to use to control for PS, component analysis to correct for genetic “clusters” even with this crude approximation we found effects that are unevenly distributed across subjects (Price that vary widely across ethnic groups, providing initial et al. 2010). In the absence of such analysis, the pos- evidence of PS. sibility that seemingly strong associations are in fact Asecondstatisticalstrategyforinvestigatingthe driven by PS can go undetected (Hao et al. 2010). For likely influence of PS is to examine the effect of traits example, Figure 1, taken from a recent study of the ge- on voting that are likely stratified by ethnic ancestry but netics of aging, shows how SNPs that initially appeared should have no relation to voting. If, when interacted highly correlated with longevity display no association with church attendance, these traits also predict voting in the same manner as (L)5HTT, this would provide strong additional evidence that the (L)5HTT∗Church “never,” “at least a few times but no more than once a month,” and Attendance-voting association is an artifact of PS. In “more than once a month.” Table 2, we show that these associations are indeed

2 American Political Science Review

FIGURE 1. SNP’s that initially appear highly correlated with longevity display no association once population stratiﬁcation is controlled for

Imputed lifespan: 432 males from original cohort Imputed lifespan: 879 females from original cohort controlling for sex, smoking, and birth cohort without PCAs Controlling for sex, smoking, and birth cohort without PCAs

Imputedlifespan:Imputedlifespan: 432 malesmales fromfrom originaloriginal ccohortohort

CControllingontrolling fforor ssex,ex, ssm smokingmoking and bbirthirth ccohortohort wwithoutithout OOCAsCAs

Observed –log 10(p) –log Observed Observed –log 10(p) –log Observed

Expected –log 10(p) Expected –log 10(p)

Imputed lifespan: 432 males from original cohort Imputed lifespan: 879 females from original cohort controlling for sex, smoking, and birth cohort and 20 PCAs Controlling for sex, smoking, and birth cohort and 20 PCAs

Observed –log 10(p) Observed –log 10(p) –log Observed

Expected –log 10(p) Expected –log 10(p)

Reprinted with permission from Anatoliy Yashin, “Accumulation from Genetics of Exceptional Lifespan: Testing Mutation Hypothesis of Aging Using GWAS,” 2011. widespread. In addition to (L)5HTT, the following friends, and being judged “attractive” or “very attrac- traits each predicted voting when interacted with tive” by the survey administrator.9 We of course do church attendance: having brown eyes, wearing not take this as a reason to advocate for new fields glasses,5 poor hearing,6 having visited a dentist in the of “opto-politics,” “audio-politics,” “dental-politics,” past year,7 being diagnosed with epilepsy,8 speaking a “lingua-politics,” and “attractiveness-politics.” Nor do language other than English at home and with close we think these associations represent progress in our understanding of voting. Rather, they are precisely the kinds of spurious associations one would expect in the 5 Non-Hispanic blacks and Mexican Americans have a higher preva- presence of PS. lence of vision impairment than non-Hispanic whites. Centers We also investigated a converse phenomenon: for Disease Control and Prevention National Center for Health Statistics: Vision, Hearing, Balance, and Sensory Impairment in whether (L)5HTT predicts voting when interacted with Americans Aged 70 Years and Over: United States, 1999–2006. other social practices that are consistent with FD’s http://www.cdc.gov/nchs/data/databriefs/db31.htm theory, but are unlikely to be ethnically stratified in 6 Non-Hispanic blacks and Mexican Americans have a higher the same manner as church attendance. Recall that prevalence of hearing problems than non-Hispanic whites. Vision, Hearing, Balance, and Sensory Impairment in Americans Aged FD’s underlying hypothesis is that (L)5HTT exercises 70 Years and Over: United States, 1999–2006.http://www.cdc.gov/ an influence on voting, conditional on church atten- nchs/data/databriefs/db31.htm dance. FD (2008, 583) begin with the uncontroversial 7 Rates of tooth decay and periodontal disease can be linked to assumption that “religious groups build a sense of be- ethnicity and country of origin even among immigrants who have longing to a larger community,” which explains why lived for many years in the United States and have increased income and education levels (“Study Suggests Link” 2007). 8 For ethnic differences in rates of epilepsy, see DeLorenzo et al. 9 Forthe influence ofethnic characteristics onestimations ofphysical (1996) and Theodore et al. (2006). attractiveness, see, e.g., Maddox (2004).

3 Genopolitics and the Science of Genetics May 2013

behavior. In describing the relationship between high TABLE 1. (L)5HTT x Attend Association school community service and voting, Hart et. al. (2007, with Voter Turnout in 10 Largest Ancestry 213) comment, “The most striking finding to emerge Groups from our study is that high school community service Country of Origin N Coef se p predicted adult voting and volunteering, after controlling for other relevant predictors and demographic Germany 1289 0.10 0.21 0.64 variables.” Lopez and Moore (2006), using data from Ireland 729− 0.22 0.29 0.44 the National Youth Survey of Civic Engagement, re- Mexico 654 0.36 0.27 0.18 port that high school students who engaged in extra- Italy 536 0.48 0.33 0.14 mural sports were 15% more likely to be registered to England 4830.58 0.35 0.10 vote and 9.4% more likely to have voted in the pres- Philippines 274 0.94 0.43 0.03 − idential election in 2000. According to Frisco, Muller, Poland 223 0.16 0.55 0.77 and Dodson (2004, 673), “Participation in scouts, reli- Scotland 191 0.17 0.52 0.74 gious youth groups, non-school team sports, and 4-H 168 −0.95 0.59 0.10 France positively predicts young adults’ voter-registration sta- Cuba 167 −0.57 0.68 0.40 Iceland & N. Ireland 127 −0.01 0.65 0.99 tus, and scouting, religious youth group membership, China 99 −0.63 0.64 0.32 and leadership positions in voluntary organizations are − positively related to voting in a presidential election.” As the columns of Table 3 under the heading “Ac- tivity Variables Tested Alone” show, each of these variables by itself is indeed associated with higher religiosity reliably predicts voter turnout. They then hy- voter turnout. However, as the columns under the pothesize that long alleles of 5HTT enhance (or short heading “(L)5HTT Interactions” show, none of these alleles suppress) the “pro-social” effect of religious same variables predict voter turnout when interacted practice, even though these alleles apparently exercise with (L)5HTT.Thisfindingconstitutesdirectevidence no direct effect on voting within the population at large. against FD’s larger theory. It also supports our larger In addition to religion, this (L)5HTT “enhancing” claim concerning population stratification to the extent effect should apply to other pro-social, community- that these activities are unlikely to be stratified in the building practices that are known to be associated with same way or to the same degree as religious attendance increased voter turnout. Add Health contains data on (which could mean either more or less stratified). This the following activities, which could each plausibly be would explain why they exhibit no interactive effect construed as activities that foster community building with (L)5HTT, contrary to the empirical implications and pro-social behavior: how often one played a team of FD’s theory. sport in the last week, how often one “hung out” with friends in the last week, how many hours a week were spent at school (which would encompass involvement Omitted Variable Bias in afterschool activities such as school clubs), whether one attended a political rally or march in the last year, Although population stratification is likely the primary and whether one volunteered or did community service cause of the correlation between voting and the in- in the last year and with which groups (we examined teraction of (L)5HTT with church attendance, FD’s the four most frequently indicated groups: youth orga- results also suffer from omitted variable bias. Political nizations such as the scouts, community centers/social scientists have studied voting behavior for a long time action groups, church-related groups, and educational and identified a host of factors that influence turnout. associations). Chief among these are three factors that Deppe et al. It is important to note what studies have already include as controls in their experimental analysis: ed- shown concerning the relationship between participa- ucation (Nie, Junn, and Barry 1996), income (Brooks tion in such activities during high school and voting and Brady 1999), and partisanship (Bartels 2000). In

TABLE 2. Traits that predict voting when interacted with religious attendance

Coef se p

Attend (L)5HTT (LL,Ls) 0.1860.0750.014∗∗ × Attend BrownEyes 0.129 0.061 0.034∗∗ × − Attend Wear Eyeglasses 0.157 0.061 0.011∗∗ × − Attend Poor Hearing 0.268 0.125 0.032∗∗ × − Attend Visited Dentist 0.212 0.061 0.001∗∗∗ × − Attend Epilepsy 0.623 0.250 0.013∗∗ × − Attend Speak Foreign Language at Home & w/Friends 0.310 0.112 0.006∗∗∗ × − Attend Rated Attractive/ Very Attractive by Interviewer 0.132 0.060 0.028∗∗ × −

4 American Political Science Review

TABLE 3. “Pro-social” behaviors that show a direct association with voting show no association when interacted with (L)5HTT.

Activity Variables (L)5HTT Interaction (with Tested Alone (w/ All Associations Tested Individually standard controls) standard controls)

DV Vote coef se p coef se p = (L)5HTT Team Sport Last Week (0,1–2,>2) 0.004 0.092 0.967 0.121 0.039 0.002 (L)5HTT × Hang Out w/FriendsLastWeek(0,1–2,>2) 0.007 0.0092 0.935 0.166 0.036 0.000 (L)5HTT × Hours at School (0,1–30, >30)− 0.079 0.103 0.439 0.538 0.042 0.000 (L)5HTT × Attend Pol Rally 0.230 0.360 0.522 1.657 0.151 0.000 (L)5HTT × Volunteer 0.054 0.121 0.655 0.910 0.051 0.000 (L)5HTT × Vol w/Youth Org 0.023 0.210 0.911 0.825 0.085 0.000 (L)5HTT × Vol w/Community/Social Action Group −0.044 0.192 0.818 0.655 0.079 0.000 (L)5HTT × Vol w/Church Group0− .034 0.178 0.847 0.832 0.074 0.000 (L)5HTT × Vol w/Educational Association 0.158 0.212 0.456 0.913 0.086 0.000 N 9,247× = addition, family background (de Vries, de Graf, and measure of church attendance, we found that FD’s re- Eisinga 2009), residence history (McNulty, Dowling, sults are tenuous, to say the least. Table 5 shows that and Ariotti 2009), and altruism (Julio 2009) have when (L)5HTT is interacted with this new measure of all been shown to influence voter turnout. Table 4, church attendance it is not remotely correlated with column 1, shows that when we include these addi- voting frequency, regardless of whether one uses FD’s tional factors as controls, using a number of variables coding or the more biologically realistic triallelic coding available in the Add Health data, FD’s result is no (coef .02, p .61 in the former case, and coef longer significant (coef .12, p .14; variable defi- .01,= ∼p .77= in∼ the latter; ordered logit models nitions are contained in= the supplemental= Online Ap- yield= ∼ the same= ∼ substantial results). Finally, columns pendix). The online appendix can be found at http:// 3–6 of Table 4 show that when we interact (L)5HTT www.journals.cambridge.org/psr2013011. with Wave IV church attendance data to examine the Note that the conditional effects of these con- association with voting, the results are not particularly trol variables are all significant. As these controls impressive (coef .13, p .09 with FD’s coding and are taken into account they displace the effects of controls, and coef= .1, p= .12 with triallelic coding = = (L)5HTT∗Church Attendance. This finding holds when and controls). we expand the analysis to include the Wave I of Add Health data as well (coef .13, p .09), as shown in column 2 of Table 4. In= examining= this combined Deppe, Stoltenberg, Smith, and Hibbing’s dataset, it is important to control for the high degree Findings of family relatedness exhibited in the first sample. We As Deppe et al. note, they are unable to replicate FD’s followed the procedure we advocated in our original finding of an effect of 5HTT∗ Church Attendance on article: repeatedly sampling one individual from each voting behavior. The only effect Deppe et al. do find family ID and averaging the distribution of effects over comes from using FD’s original 5HTT coding and a 500 samples. Incidentally, FD mischaracterize our ap- general measure of political participation that they proach and conduct only a single, random draw in their have constructed based on subject responses to six analysis, but this approach is indefensible given the questions concerning political participation. As they large variance of effects revealed by repeated draws. note, the association they find is tenuous and decreases Finally, we wish to draw attention to a revealing in significance when they discard the small number of fragility in the (L)5HTT∗Church Attendance result nonwhite study participants. Nor are they able to dis- that FD do not note. In response to concerns we raised cern an effect on either voting behavior (using mea- about the definition of “voter turnout” (i.e., whether sures of validated voting) or political participation voting for the first time between ages 18–22 in the 2000 when using what is believed to be the most biolog- presidential election is a good indicator of whether one ically sound categorization of long 5HTT (triallelic remains a “voter”), FD examine a variable from Wave coding): IV of Add Health that asks respondents to indicate how often they usually vote in local or statewide elections. With this triallelic genotype classification, the most ac- They show that this measure of voter turnout continues curate according to the latest research, 5-HTT genotype to be (weakly) associated with (L)5HTT∗Church At- continues to exhibit no statistically significant relationship tendance. However, they neglect to mention that Wave with self-reported political participation. When we ran this IV of Add Health also (again) asks participants how same model with voting frequency as the dependent vari- often they attend church services. Using this second able the relationship even gave some indication of going

5 Genopolitics and the Science of Genetics May 2013 0.050 0.002 0.000 0.000 8 8 8 8 0.015 0.000 0.154 0.614 0.049 0.000 8 8 90.0540.000 8 8 103 0.067 0.124 634 0.361 0.000 094 0.04 014 0.053233 0.790 0.057 0.000 335 0.10 035 0.075 0.643 11 595 0.06207 0.000 170 0.070 0.016 206 0.04 113 0.050 0.024 6 193 0.050 0.000 62 317 0.050 0.000 000 0.000 0.053 246 0.159 0.000 469 0.05 306 0.123 0.013 271 0.062 0.000 ...... 4 0 0 0 0 0 ,727 − − − − − − 8 0.000 0.027 0 e4AttendData v 8 8 0.063 0.117 0 8 40.1430.007 8 9 09 202 0.27 314 0.095 0.001 035 0.049312 0.479 0.053 0.000 0 149 0.06 100 0.013 0.000 0 107 0.044 0.015 209 0.0543 0.000 0 ...... 88 0 0 0 0 0 2 , − − − − − 8 − ith Voting Using Wa w 0.050 0.000 0 0.003 0.000 0 8 8 8 8 0.015 0.000 0 0.049 0.000 0 8 003.0 0 70.0630.000 90.0540.000 0 8 8 8 129 0.077 0.093 0 031 0.061203 0.606 0.069 0.003 0 171 0.070 0.015 037 0.075 0.625 11 094 0.04 206 0.04 5 074 0.154 0.631 112 0.050 0.025 0 6 193 0.050 0.000 0 321 0.10 62 316 0.050 0.000000 0.000 0.053246 0.159664 0.000 0.362 0.000 0 0 1 467 0.05 304 0.123 0.014 0 271 0.062 0.000 0 ...... 0 0 0 0 4 ,727 − − − 8 − − Attend Association ∗ 8 03 0 8 0.030 0 8 10.1430.00 8 9 141 0.072 0.049 0 014 0.057 0. 269 0.065 0.000 0 147 0.06 100 0.013 0.000 0 3 107 0.044 0.015 203 0.054 0.000 0 295 0.095 0.002 239 0.279 0.000 ...... 88 0 0 0 0 2 , − − − 8 − − 0.000 0 0.050 0.000 0 0.003 8 8 8 8 ined b 0.069 0.003 0 0.10 8 60.04 40.0490.000 0 30.0700.015 Com 8 8 88 8 131 0.077 0.093 0 009 0.061 0.606 0 26 1 097 0.050 0.025 0 009 0.075 0.625 120 0.015 0.000 0 036 0.154 0.631 070 0.04 652 0.054 0.000176 0.050 0.000 0 0 575 0.063 0.000 0 5 277 0.050 0.000000 0.000 0.053232 0.159 0.000 0 0 1 433 0.05 2 326 0.123 0.014590 0.362 0.000 0 253 0.062 0.000 0 1 14 14 9,247 9,066 9,247 9,066 ...... 8 8 0 0 0 0 0 4 0 Independent Sample FD’s 5HTT Coding (LL,Ls) Triallelic 5HTT Coding − − − − − − 8 0.000 0 0.046 0.000 0 0.004 0150 00.135 8 8 8 8 8 Sample 0.072 0.000 0 0.015 0.000 0 0.054 0.000 0 0.159 0.000 1 Common Predictors of Voting 5HTT w 8 8 000.0 0 40.0500.000 8 8 8 44 0.364 0.000 Ne 120 0.0 030 0.061 0.627 26 196 0.04 110 0.050 0.02 029 0.075 0.701 0 12 067 0.155 0.667 096 0.04 1 66 597 0.062 0.000 0 626 0.049 0.000 0 000 0.000 0.04923 0 307 0.050 0.000 0 451 0.05 316 0.10 309 0.124 0.012 0 8 254 0.062 0.000 0 167 0.070 0.017 ...... 0 0 0 0 4 0 ,727 9, Association of (L)5HTTxAttend, Controlling for 8 − − − − − w nt 0 u 25K 0 rn 0 u > als 9,066 9, u id Attend(W3/4) 0 ith Parents 0 Vote coef se p coef se p coef se p coef se p coef se p coef se p v ∗ e American w cation (college) 0 v eEmailAcco = v eatSameAdd. 0 e u v v DV 5HTT TABLE 4. 5HTTxAttend Association with Voting Using Controls, Full Sample, Wave 4 Attend Data, and Alternate 5HTT Coding. 5HTT 0 NFamilies Attend(W3/4)Age 0 0 Li Hispanic 0 N Indi Male Parent Income Black 0 Li Nati Ed Income 0 Organ Donor 0 Asian Identify Political Vie Attended Pol. Rally 1 Ha Constant Born in the US 0 Filed a Tax Ret Father in Jail

6 American Political Science Review 0.000 ency u 00.001 8 8 016020 0.055 0.769 0.043 0.647 449 0.046 0.000 445 0.126 0.000 072 0.011 0.000 272 0.0 234 0.03 219 0.059 0.000 322 0.047 0.000 ...... ith Voting Freq 0 0 0 0 0 0 w − − − − − Coding 8 e 4 Attend Data and Triallelic 5HTT v 0.037 0.000 0.149 0.000 Attend Association ∗ Regression 4 Ordered Logit 8 8 3 Mixed Effects 49 010 0.034 0.765 012 0.027274 0.66 0.029 0.000 0 230 0.076 0.002 044 0.007 0.000 0 193 0.051 0.000 144 0.024 0.000 13 50 205 0.030 0.000 0 Using Wa ...... 8 5HTT , 0 0 0 0 0 8 − − − − − 8 00 8 0.000 ency u 8 00.001 8 030 0.063 0.62 413 0.056 0.000 0 444 0.126 0.000 066 0.049 0.1 072 0.011 0.000 0 234 0.03 260 0.0 215 0.059 0.000 314 0.047 0.000 0 ...... ith Voting Freq 0 0 0 0 w − − − − e4AttendDataandFD’s5HTTCoding v Attend Association ∗ Regression 2 Ordered Logit 60.0510.000 90.1500.000 1 8 8 1 Mixed Effects 49 229 0.076 0.003 020 0.039 0.613251 0.035 0 0.000 0 037 0.031 0.235 0 044 0.007143 0.000 0.024 0.000 0 1 4 136 0.037 0.000 201 0.030 0.000 0 ...... 8 , 0 0 0 0 0 0 0 1 0 0 5HTT 8 9,201 9,201 9,201 9,201 Using Wa − − − − ency coef se p coef se p coef se p coef coef se p u AttendW4 Vote Freq ∗ = AttendW4 TABLE 5.Coding. 5HTTxAttend Association with Vote Frequency Using Wave 4 Attend Data andDV Alternate 5HTT 5HTT 5HTT Age Male Asian Constant Hispanic NFamilies N Inviduals Black Nativeamer

7 Genopolitics and the Science of Genetics May 2013

in the opposite direction from that hypothesized by Fowler the existence of something called an “endophenotype.” and Dawes though the coefficient was not significant (13). We strongly encourage readers to look at an expanded (emphasis added) table of claimed associations for the same polymorphic regions of the same four genes (MAOA, 5HTT, DRD2, Thus Deppe et al.’s experimental evidence does DRD4)10 and consider whether proposing “several en- not confirm FD’s voter turnout thesis, and one won- dophenotypes” is a convincing explanation. ders how many specifications of “political participa- Concerning the hundreds of associations reported tion” were explored before arriving at the one that between the same polymorphic regions of the same shows a weak association with FD’s original coding four genes and every imaginable behavior (as well as (presumably, if any of the six components used to nonbehavioral phenotypes), FD comment that “many construct the participation measure showed an indi- of the phenotypes listed by CE have not been repli- vidual association with (L)5HTT∗Church Attendance, cated, so it may be premature to ask what such differ- that would have been reported as well). Given that ent phenotypes have in common with one another.” genetic stratification can still be a significant prob- In fact, as we noted (Charney and English 2012, 11), lem within racially homogeneous groups (particularly the most well-known associations on this list—that be- “Caucasians” at large), it is reasonable to think that tween MAOA and “antisocial” personality and be- population stratification is driving the weak association tween 5HTT and depression—which are typically refer- they do find (Price et al. 2009). Overall, we view Deppe enced as scientific facts, have failed to be replicated as et al.’s results as constituting more of a challenge to many times as they have been replicated. One wonders than a confirmation of FD’s thesis. then, if such associations can ever be proven wrong; that is, are they falsifiable and hence, truly scientific claims? GENETICS (Popper 2002). And why has the canonical test of scientific validity—consistent replication—beensuspended in Responses to a Few Objections these cases? FD insist that they “clearly acknowledge that there FD note that 40 polymorphisms have been reliably as- are multiple genetic and environmental causal factors sociated with type 1 diabetes (T1D) and 50 with type that underlie turnout” by quoting their 2008 article 2(T2D).Thisistrue.However,inthecaseofT2D, (Fowler and Dawes 2008, 590): “[T]here is some (likely for example, it is estimated that the existing 50 genetic large) set of genes whose expression, in combination markers only explain 15% of the heritability of the with environmental factors, influences political par- disease, and in neither T1D nor T2D do the associated ticipation.” Indeed, a recent study identified up- and markers have any predictive (i.e., diagnostic) value: downregulation of more than 4,038 genes in differences Only a small proportion of those who are considered in aggression in fruit flies (Zwarts et al. 2011). It is genetically susceptible ever develop the disease, indi- not sufficient, however, to pay lip service to the exis- cating the critical importance of environmental factors tence of such extreme polygenicity and environmental (Knip et al. 2012). This example supports rather than interaction and then defend the plausibility of results— challenges our point. If such is the level of complex- that a single polymorphism can predict a behavior such ity for a well-defined disease such as diabetes, are we as voter turnout—that directly contradict those same to believe that a single polymorphism predicts voting, principles. aprocessthatinvolvesallofthefacultiesofthehu- man brain (consciousness, thinking, reasoning, mem- Transcribability, Transcription, and ory, planning, emotion, etc.) interacting with a partic- Translation ular environment? Moreover, whether polymorphisms will ever be discovered that can predict the onset of FD assume that, on the basis of our claim that the T1D or T2D has little to do with the question whether presence of a particular polymorphism alone cannot asinglepolymorphismcanpredictacomplexbehavior tell us its epigenetic state, we infer that “the gene plays such as voting. We invoked the example of diseases no role in influencing behavior,” which “demonstrates such as T1D simply to illustrate how hard it is to identify aprofoundmisunderstandingofscientificinference.” risk-factor polymorphisms in cases where the existence We nowhere make any such inference: Mutations on, of such genes makes genetic and biological sense in for example, the NOTCH3 gene result in cognitive de- the first place (for more on this point, see the later fects and dementia. Hence, a gene can play a role in discussion). influencing a particular behavior. In defending the possibility that the same polymor- FD exhibit a lack of awareness of the differences be- phisms of the same four genes could predict hundreds tween (1) the extent to which a gene can be transcribed of widely divergent behavioral and nonbehavioral phe- (its accessibility to transcription factors), (2) gene tran- notypes, FD invoke pleiotropy, the phenomenon in scription, and (3) gene translation. This is apparent which the same proteins transcribed from the same throughout their comments and is made explicit in their genes are involved in many different physiological pro- assertion that our claim that genes do not regulate the cesses. There is abundant evidence for the existence extent to which they can be transcribed is “directly of widespread pleiotropy. FD, however, conflate the notion of pleiotropy with that of an endophenotype. Unlike pleiotropy, there is no scientific evidence for 10 See http://tinyurl.com/4genes.

8 American Political Science Review contradicted by evidence from more than 5,000 genes tein abundance. Some mRNAs are translated into that shows transcription explains nearly 40% of the only one protein per hour, others are translated 200 variation in levels of expression in mammals.” In the times” (emphasis in original; “From Gene to Protein” supporting study that FD cite (Schwanhausser¨ et al. 2011). 2011), the authors report their findings on the relation It is thus critical to be aware of the following basic between gene transcription, gene translation, and the principle of genetics: Gene accessibility to transcription levels of protein synthesized in a cell. factors, gene transcription, and gene translation are Transcription is a process in which a segment of DNA three distinct processes in the pathway leading to the is copied (by the cellular machinery) to produce mes- synthesis of a protein, each subject to its own complex senger RNA (mRNA), which is then used to construct regulatory system. aparticularproteininaprocessknownastranslation. Before transcription can occur, however, the segment Shifting Paradigms and Overturned Dogmas of DNA to be transcribed must be accessible to special proteins called transcription factors. This accessibility is FD respond to our reference to retrotransposons regulated not by the gene itself but by the epigenome, (“jumping genes”) that alter both DNA sequence and the name given to a variety of complex biochemical quantity by stating that our description is “highly mis- processes that can block or facilitate the access of a seg- leading” and “extremely disingenuous” because “all ment of DNA to transcription factors (Allis, Jenuwein, but a few dozen of the 3 billion base pairs in an individ- and Reinberg 2007). ual’s DNA will be exactly the same throughout their The measure used by Schwanhausser¨ et al. (2011) of reproductive lifetimes.Thus,byandlargethegenes the relationship between the efficiency of transcription you are born with are the genes you will die with” (as measured by levels of mRNA) and intracellular (reference omitted). protein levels applies when transcription is actually oc- Retrotransposons are part of what we, following a curring. If a gene is inaccessible to transcription factors, number of prominent geneticists and neuroscientists, then transcription cannot occur, no matter what the re- referred to as a “paradigm shift” in genetics: lationship between transcriptional efficiency and cellular protein levels is when the gene is being transcribed. Today, 50 years after these events took place [after the Transcriptional efficiency has no bearing on epigenetic discovery of transposable elements], nobody would deny that genomes contain a wealth of DNA sequences able regulation of whether or not a gene can be transcribed to move, usually referred to as transposition, from one in the first place. Let us make this perfectly clear: The genome site to another, using different mechanisms. But same 5HTT gene that produces serotonin transporter this consensus was difficult to reach because it repre- in neurons is also present in eye and heart and hair cells. sented a paradigm shift in theories of genome stabil- We cannot predict the levels of serotonin transporter ity and control. Transposable elements are now incor- in hair cells on the basis of a person’s 5HTT genotype porated into the contemporary concept of the genome because, in hair cells, the 5HTT gene is not transcribed as an entity with unsuspected dynamism and fluidity of at all.Itispermanentlyepigeneticallysilenced,some- far reaching evolutionary consequences (Fontdevila 2011, thing that the presence of the 5HTT gene alone cannot p. 81). tell us (Bird 2007; Khavari, Sen, and Rinn 2010). Thus, Novel classes of small and long noncoding RNAs (ncR- to repeat what we said previously, genes do not reg- NAs) are being characterized at a rapid pace, driven ulate the extent to which they are capable of being by recent paradigm shifts in our understanding of ge- transcribed in any obvious, unidirectional manner. nomic architecture, regulation, and transcriptional out- Significantly, the takeaway conclusion of the study put, as well as by innovations in sequencing technologies of Schwanhausser¨ et al. (2011) that FD cite concerns and computational and systems biology. These ncRNAs not transcription but translation, and the authors’ con- can interact with DNA, RNA, and protein molecules; clusion challenges the very claim that FD invoke this engage in diverse structural, functional, and regulatory activities; and play roles in nuclear organization and study to defend. What Schwanhausser¨ et al. (2011, 337, transcriptional, post-transcriptional, and epigenetic pro- 341) highlight as the central finding of their study is cesses. This expanding inventory of ncRNAs is impli- that translational efficiency is a far better indicator of cated in mediating a broad spectrum of processes in- protein synthesis than transcriptional efficiency and is cluding brain evolution, development, synaptic plastic- the key determinant of intracellular protein levels: ity, and disease pathogenesis (Qureshi and Mehler 2012, 528). We find that cellular abundance of proteins is primarily Epigenetics provides an additional molecular mechanism at the level of translation.... Hence, protein abundance to complement genetics in the regulation of development. seems to be predominantly regulated at the ribosome Therefore, the paradigm shift is that layers of molecular [structures in a cell where messenger RNAs are trans- control and cascades of both epigenetic and genetic fac- lated to construct proteins], highlighting the importance tors or processes are involved in regulating developmental of translational control....Wefoundthatinmousefibrob- biology (Skinner 2011, 52) lasts, translation efficiency is the single best predictor of protein levels. (emphasis added) However, FD dismiss the idea of a paradigm shift: “Thus the “paradigm” is not at issue – it is the As one of the authors notes in another article, “The methodology that proves challenging.” Apparently, ribosomes [ translation] ultimately determine pro- the paradigm does not change (hence, discoveries in =

9 Genopolitics and the Science of Genetics May 2013 molecular genetics over the past 50 years can be ig- These changes to DNA are somatic;thatis,theyare nored), and the real challenge lies in the methodology changes to DNA that occur postconception in the em- (i.e., statistical analysis). Because of the significance of bryo and can affect all the cells of the body except germ retrotransposons in this paradigm shift, we consider cells (i.e., egg and sperm—the “germline”). As such, them here at length, together with another source of these changes are not transmitted to offspring, in con- DNA variability. We also explain the basic difference trast to changes in germline DNA (Notini, Craig, and between somatic and germline DNA mutability, a dis- White 2008). Over the past 20 years, researchers have tinction FD are apparently unaware of, and its rele- come to appreciate both the extent and the importance vance in this context. of somatic DNA mutability for human phenotypes (De Active (or “transpositionally competent”) retro- 2011). transposons are segments of DNA that move about the Retrotransposition is only one source of postconcep- genome by a “copy and paste” mechanism. They first tion DNA variability. Another source is aneuploidy, the copy themselves to RNA, and the original DNA copy presence of greater or fewer than two chromosomes— is maintained at the same location. The RNA copy is and hence two alleles of each gene—per cell. Recent then “reverse-transcribed” into DNA, and the DNA is conservative estimates place the overall percentage inserted into the genome at a new location (Sciamanna of aneuploid neural cells—cells that vary from the et al. 2009). Hence, these elements expand in number as usual two chromosomes per cell—in the normal human they retrotranspose, leading to an increase in genomic brain at an astonishing 10%, involving monosomy (one DNA content and a change in DNA sequence and chromosome), trisomy (three chromosomes), tetra- structure at the region of insertion (Faulkner 2011). somy (four chromosomes), polyploidy (>four chro- For the most part, the activity of retrotransposons is mosomes), and uniparental disomy (two copies of a epigenetically silenced due to potentially deleterious chromosome from one parent; Rehen 2005). Given effects, but there are two critical exceptions: First, for 100 billion neurons in the adult brain, this yields a aperiodduringearlyembryogenesis,retrotransposons rough∼ conservative estimate of 10 billion aneuploid are released from epigenetic suppression and become neurons. This chromosomal diversity appears to result active (Coufal et al. 2009) and may influence the man- from a high frequency of stochastic errors in cellular ner in which neural precursor cells differentiate to form division during embryogenesis, and various lines of ev- distinct types of neurons in the embryo (Vitullo et al. idence indicate that brain tissues may be more prone 2012); second, retrotransposition is ongoing in those to aneuploidy than other tissues (Iourov, Vorsanova, parts of the brain (the hippocampus and the subven- and Yurov 2006). Mature aneuploid neurons are func- tricular zone) that produce new neurons throughout tionally active and integrated into brain circuitry, show- life (Muotri et al. 2010). ing distant axonal connections (Kingsbury 2005). One Applying high-throughput sequencing techniques likely result of this integration is neuronal signaling to study retrotransposition in human brain tissue, differences caused by altered gene expression, as doc- Baillie et al. (2011) identified 25,000 retrotransposon umented in mammalian neural cells (Kaushal et al. insertions in the hippocampus∼ and caudate nucleus of 2003). Thus, a network composed of intermixed diploid healthy individuals. A number of key genomic loci were and aneuploid neurons might produce unique signaling found to contain these insertions, including dopamine properties distinct from a network composed purely of receptors and serotonin neurotransmitter transporters. diploid cells (Westra et al. 2010). They also identified a disproportionate number of in- Therefore, the assertion that “all but a few dozen tronic retrotransposon insertions, which is noteworthy of the 3 billion base pairs in an individual’s DNA because introns are the protein-coding loci of DNA; will be exactly the same throughout their reproduc- they also determined that genes containing intronic tive lifetimes” is simply wrong. Pervasive changes to a insertions were twice as likely to be differentially over- person’s genome beginning with conception and con- expressed (i.e., overtranscribed) in the brain. What is tinuing throughout life pose a significant challenge to entailed by ongoing retrotransposition in the brain is any methodology that rests on the assumption that the summed up by the title of the study of Baillie et al. genome is the unchanging, static template of heredity, (2011): “Somatic retrotransposition alters the genetic identical in all the cells and tissues of the body. We landscape of the human brain” (for more on retro- cannot assume that any two individuals have two copies transposons and alteration to neuronal DNA, see, e.g., of, say, 5HTT alleles in all their neurons or that these Coufal et al. 2009; Faulkner 2011; Fontdevila 2011; copies do not exhibit transcriptional differences due Muotri et al. 2010; Sciammana et al. 2009; Singer et al. to the activity of retrotransposons or to environmental 2010; Vitullo et al. 2012). Given that each retrotrans- reprogramming of the epigenome (see the later discus- position event that occurs in a cell results both in an sion). Moreover, although we characterized both retro- increase in DNA and a change in the DNA sequence, transposition and aneuploidy as postconception so- the discovery of different numbers of retrotransposi- matic occurrences, both retrotransposons (Sciamanna tion insertions in different neurons indicates neuronal et al. 2009) and aneuploidy (Delhanty 2011) can be somatic mosaicism,theexistenceinthesameindivid- inherited via the germline as well. Ongoing germline ual of two or more distinct genomes: “[G]enome mo- retrotransposition is now believed to have played a saicism driven by retrotransposition may reshape the key role in human evolution (Iskow et al. 2010). genetic circuitry that underpins normal and abnormal Another reason we cannot assume that there are neurobiological processes” (Baillie et al. 2011, 534). only two copies of an allele per cell in any given

10 American Political Science Review individual (and any given cell) is the ubiquitousness but an environment as well. Behavioral developmen- of copy number variations (CNVs)—stretches of DNA tal plasticity evolved because it is adaptive,promoting at least 1,000 base pairs (1 kilobase) long and extend- Darwinian fitness by enhancing survival and reproduc- ing up to several million base pairs—that are either tive success through the use of environmental cues deleted or are present in multiple copies relative to a to optimize the life-course (Garland and Kelly 2006). model normal genome (Dear 2009). A first-generation Numerous animal studies have shown that one of the CNV map of the human genome showed that at least ways in which the perinatal environment can shape 2,900 genes, or 10% of the total number of genes in the behavioral phenotypic outcomes (e.g., stress responses, human genome, contain or are encompassed by CNVs aggression, mating behavior) to meet the demands of (Redon et al. 2006). The average size of the CNVs was the postnatal environment is via environmental repro- 250,000 base pairs. Since the average gene is 27,000– gramming of the epigenome, resulting in long-term 60,000 base pairs long, many of the CNVs were com- changes in gene transcribability (Champagne 2010; Fa- posed of multiple copies (or deletions) of entire genes, giolini, Jensen, and Champagne 2009; Mychasiuk et al. in some cases exceeding 12 copies of a single gene. 2012; Zhang and Meaney 2010). Furthermore, as do retrotransposons and aneuploidy, Absurdly high estimates of heritability of behavior CNVs contribute to somatic mosaicism (i.e., different (of the kind typically obtained by classical twin studies) CNVs have been found in different cells and tissues are incompatible with phenotypic plasticity. Were we to of the same individual). As Singer et al. (2010, 345) take an estimate of, say, 69% heritability of aggression note, “Neuronal genetic diversity results from aneu- in humans seriously (versus 10% in Drosophila), then ploidy (whole chromosome gains and losses) genomic we would have to conclude that, even though humans copy number variations (CNVs) and actively ‘jumping’ possess the most plastic, responsive, adaptive organ we transposable elements” (references omitted). know of in the natural world (the human brain), they are less developmentally influenced by and responsive to their environment than flies. Lacking the ability to adapt to their environments to such an extent, homo Neuroscience: Plasticity, Criticality, and 11 Adaptation sapiens would long ago have become extinct. Plasticity is built into the neurodynamics of the hu- Whole-genome expression profiling has identified dif- man brain. There is a good deal of hard scientific evi- ferences in the transcription levels of more than dence from cutting-edge research in neurobiology that 4,038 genes in hyperaggressive fruit flies (Drosophila macroscopic behaviors (cognitive, emotional, motor, melanogaster)versuscontrols(Zwartsetal.2011).For etc.) are emergent phenomena of an underlying neu- all of the up and down transcription levels in thousands ronal collective characterized by self-organized crit- of genes, the heritability of Drosophila aggression is icality (Chialvo 2010; Droste, Do, and Gross 2013; estimated to be only 10%, even though the researchers Proekt et al. 2012). Emergence refers to the unexpected thought they had raised all the flies in identical labora- collective spatiotemporal patterns exhibited by large, tory conditions (Edwards et al. 2006). Why were the complex systems, where “unexpected” indicates our heritability estimates so low compared to estimates of inability (mathematical and otherwise) to derive such 69% for the heritability of aggression in humans (van emergent patterns from the equations describing the den Oord et al. 1996), 50% for the heritability of po- dynamics of the individual parts of the system. Com- litical ideology (Alford, Funk, and Hibbing 2005), and plex systems (such as the brain) are usually large 65% for the heritability of being a born-again Christian conglomerates of interacting elements, each one ex- (Bradshaw and Ellison 2008)? First, because they were hibiting some sort of nonlinear dynamics (Chialvo accurate,involvinggeneticandenvironmentalmanipu- 2010). The essence of self-organization is that a sys- lation, continual monitoring, and whole-brain analysis tem structure (at least in part) appears without ex- of healthy samples—conditions obviously impossible in plicit pressure or constraints from outside the sys- any human study. Second, aggression is a highly adap- tem. Criticality is a mathematically defined complex tive behavior. state at the border between predictable period behav- Phenotypic plasticity is the ability of an organism ior and unpredictable chaos. In fMRI experiments, it to change phenotype in response to its environment has been demonstrated that functional neuronal net- (Pigliucci, Murren, and Schlichting 2006). It includes works exhibit scale invariance,12 afeatureofcriticality the possibility of modifying developmental trajectories (Eguıluz´ et al. 2005; He et al. 2010; Plenz 2012). In the in response to specific environmental cues and the ability to change phenotypic state or activity in response 11 For an extended treatment of these topics, see Charney (2012). to variations in environmental conditions (both pre- 12 Scale-invariant phenomena exist in the vicinity of a continuous and postnatal). Modern evolutionary biology reflects phase transition where processes at the microscopic, macroscopic, the idea that adaptation is not limited to the process and indeed all intermediate scales are essentially similar except for of natural selection (i.e., adaptation at the level of the achangeinscale.Moreformally,afunctionf(x)issaidtobescale- invariant if, on multiplying the argument of the function by some species), but also includes adaptation of the individual (β 1) organism to its ecological niche (West-Eberhard 2003). constant scaling factor (λ), one obtains f(λx) λ− + f(x): The same shape is retained but with a different scale.= It is straightforward Offspring do not inherit simply genes (and messen- to show that a function that satisfies this property is a power law (β 1) ger RNA, noncoding RNA, epigenomes, mitochondria, p(x) < x− + ,where(β 1) is the scaling exponent (Proekt et al. mitochondrial DNA, and nucleoli) from their parents 2012). +

11 Genopolitics and the Science of Genetics May 2013

context of the dynamics of the brain, the implication simple regression analysis, it does not exist. The spec- of self-organized criticality is that the cerebral cortex tacular advances in our understanding of the genome displays perpetual state transitions, dynamics that favor over the last several decades pose a direct challenge to reacting to inputs quickly and flexibly (Shew and Plenz the simplistic model of the genome and the genotype- 2012). Why should our brains evolve to be near a crit- phenotype relationship on which genopolitics relies: ical point? “The answer, in short, is that brains should The genome is not the unchanging template of hered- be critical because the world in which they must survive ity, fixed for life at the moment that the maternal and is to some degree critical as well” (Chialvo 2010, 747). paternal chromosomes fuse in the fertilized egg cell If complex behaviors associated with the healthy and identical in all the cells and tissues of the body; it is brain are emergent phenomena of an underlying neu- not the sole biological component of inheritance; and ronal collective, then they are not the sort of thing that it is not a “self-activating,” “self-determining” “agent” can be predicted by 1 gene or 10,000 genes. When it of either protein production or phenotype creation. comes to such behaviors, genes are the wrong level of Genopolitics relies on a conception of the human analysis.Assumingthatwecanpredictcomplexbehav- brain that complements its conception of the genome. iors from genes alone (and skip everything between the For all the lip service paid to complexity, the “genopo- gene and the appearance of the behavior) is akin to as- litical brain” more resembles a mechanical toy whose suming that we can predict the tides solely by studying behavior is determined by the 25,000 little wind-up toys the molecular structure of water molecules. Note that (i.e., genes) of which it is composed than a neuronal the claim here is not that “genes do not affect behavior” collective whose behavior is characterized by emergent any more than the claim that we cannot predict the tides self-organized criticality to enable rapid and flexible by the structure of water molecules is a claim that the responses to the demands of a variable environment. atomic structure of water is irrelevant for the behavior Given that such a mechanical brain would have bro- of the tides. ken down long ago in evolutionary history, we can be To all of this it might be objected that “gene knock- thankful that it has no more reality than its mechanical out” studies—studies in which animals are engineered genome. to lack a particular gene—prove us wrong, inasmuch as they tie particular genes to particular (complex) behaviors. MAOA knock-out mice exhibit heightened REFERENCES aggression (as well as a host of other abnormal behav- Alford, J. R., C. L. Funk, and J. R. Hibbing. 2005. “Are Political Ori- iors). Does this not prove that MAOA is the “warrior entations Genetically Transmitted?” American Political Science gene” (McDermott et al. 2009)? It proves no such thing. Review 99 (2): 153–67. Angelman syndrome is a neurological disorder caused Allis, C. D., T. Jenuwein, and D. Reinberg. 2007. Epigenetics.Cold by the lack of a functioning UBEA3 gene that codes Spring Harbor, NY: Cold Spring Harbor Laboratory Press. Baillie, J. K., M. W. Barnett, K. R. Upton, D. J. Gerhardt, T. A. for an enzyme involved in the intracellular degradation Richmond, F. De Sapio, P. M. Brennan, et al. 2011. “Somatic of proteins (Clayton-Smith and Laan 2003). Hence, Retrotransposition Alters the Genetic Landscape of the Human those who have the disorder are UBEA3 “knock-outs.” Brain.” Nature 479 (7374): 534–37. Angelman syndrome is characterized by intellectual Bartels, L. M. 2000. “Partisanship and Voting Behavior, 1952–1996.” American Journal of Political Science 44 (1): 35–50. disability, severe speech impairment, and a generally Bird, A. 2007. “Perceptions of Epigenetics.” Nature 447 (7143): 396– happy demeanor with frequent outbursts of laughter, 98. but no one has concluded on this basis that UBEA3 Bouaziz, M., C. Ambroise, and M. Guedj. 2011. “Accounting is the “happiness” or “laughter” gene. Although gene for Population Stratification in Practice: A Comparison of the knock-out studies are a valuable research tool, they are Main Strategies Dedicated to Genome-Wide Association Studies.” PLoS ONE 6 (12): e28845. apt to deceive because they in effect result in an arti- Bradshaw, M., and C. G. Ellison. 2008. “Do Genetic Factors Influence ficial monogenic (i.e.,“single gene”) disorder.Voting— Religious Life? Findings from a Behavior Genetic Analysis of Twin and all other complex human behavior—is not a mono- Siblings.” Journal for the Scientific Study of Religion 47 (4): 529–44. genic disorder, or an oligogenic disorder, or a complex Brooks, C., and D. Brady. 1999. “Income, Economic Voting, and Long-Term Political Change in the U.S., 1952–1996.” Social Forces polygenic disorder. Normal human behavior is not a 77 (4): 1339–74. cluster of disorders, nor is it a cluster of distinct “be- Champagne, F. A. 2010. “Epigenetic Influence of Social Experiences haviors,” each behavior predicted by a gene or set of across the Lifespan.” Developmental Psychobiology 52 (4): 299– genes. Rather, behavior is the integrated output of an 311. integrated biological system interacting with a particu- Charney, E. 2012. “Behavior Genetics and Postgenomics.” Behav- ioral and Brain Sciences 35 (5): 331–58. lar environment. Charney, E., and W. English. 2012. “Candidate Genes and Political Behavior.” American Political Science Review 106 (1): 1–34. Chatters, L. M., R. J. Taylor, J. S. Jackson, and K. D. Lincoln. 2008. “Race and Ethnic Differences in Religious Involvement: African CONCLUSION Americans, Caribbean Blacks and Non-Hispanic Whites.” Ethnic and Racial Studies 32 (7): 1143–63. Genopolitics is an exercise in naıve¨ statistics. Genetics, Chialvo, D. R. 2010. “Emergent Complex Neural Dynamics.” Nature however, is not a subfield of statistics. Genopolitics Physics 6 (10): 744–50. relies on a naıve¨ conception of the genome uninformed Clayton-Smith, J, and L. Laan. 2003. “Angelman Syndrome: A Re- view of the Clinical and Genetic Aspects.” Journal of Medical by some basic principles of genetics and by discoveries Genetics 40 (2): 87–95. in molecular genetics over the past 50 years. Although Coufal, N. G., F. H. Gage, J. L. Garcia-Perez, M. T. Lovci, J. V. such a genome may be ideally suited for discovery via Moran, M. Morell, Y. Mu, K. S. Shea, G. E. Peng, and G. W. Yeo.

12 American Political Science Review

2009. “L1 Retrotransposition in Human Neural Progenitor Cells.” Khavari, D. A., G. L. Sen, and J. L. Rinn. 2010. “DNA Methylation Nature 460: 1127–31. and Epigenetic Control of Cellular Differentiation.” Cell Cycle 9 De, S. 2011. “Somatic Mosaicism in Healthy Human Tissues.” Trends (19): 3880–83. in Genetics 27 (6): 217–23. Kingsbury, M. A. 2005. “Aneuploid Neurons Are Functionally Ac- Dear, P. H. 2009. “Copy-number Variation: The End of the Human tive and Integrated unto Brain Circuitry.” Proceedings of the Na- Genome?” Trends in Biotechnology 27 (8): 448–54. tional Acadamy of Sciences USA 102: 6143–47. Delhanty, J. D. 2011. “Inherited Aneuploidy: Germline Mosaicism.” Knip, M., R. Veijola, S. M. Virtanen, H. Hyoty, O. Vaarala, and H. Cytogenetic Genome Research 133 (2–4): 136–40. K. Akerblom. 2012. “Environmental Triggers and Determinants DeLorenzo, R. J., W. A. Hauser, A. R. Towne, J. G. Boggs, J. M. of Type 1 Diabetes.” Diabetes 54 (Suppl 2): S125–36. Pellock, L. Penberthy, L. Garnett, C. A. Fortner, and D. Ko. 1996. Lopez, M., and K. Moore. 2006. “Participation in Sports and Civic “A Prospective, Population-based Epidemiologic Study of Sta- Engagement.” CIRCLE: The Center for Information & Research tus Epilepticus in Richmond, Virginia.” Neurology 46 (4): 1029– on Civic Learning & Engagement.http://www.civicyouth.org/ 35. PopUps/FactSheets/FS_06_Sports_and_Civic_Engagement.pdf. Devlin, B., S. A. Bacanu, and K. Roeder. 2004. “Genomic Control to Lotrich, F.E., B. G. Pollock, and R. E. Ferrell. 2003. “Serotonin Trans- the Extreme.” Nature Genetics 36 (11):1129–30; author reply, 31. porter Promoter Polymorphism in African Americans: Allele Fre- de Vries, J., N. D. de Graaf, and R. Eisinga. 2009. “Biases in the Ef- quencies and Implications for Treatment.” American Journal of fects of Family Background Characteristics on Voting Preference: Pharmacogenomics 3(2):145–47. The Dutch Case.” Electoral Studies 28 (2): 204–17. Maddox, K. B. 2004. “Perspectives on Racial Phenotypicality Bias.” Droste, F., A. Do, and T. Gross. 2013. “Analytical Investigation of Personality and Social Psychology Review 8 (4): 383–401. Self-Organized Criticality in Neural Networks.” Journal of The Marler, P.L., and C. K. Hadaway. 1999. “Testingthe Attendance Gap Royal Society Interface 10 (78): 1–8. in a Conservative Church.” Sociology of Religion 60 (2): 175–86. Edwards, A., S. M. Rollmann, T. J. Morgan, and T. F.C. Mackay. 2006. McDermott, R., D. Tingley, J. Cowden, G. Frazzetto, and D. D. P. “Quantitative Genomics of Aggressive Behavior in Drosophila Johnson. 2009. “Monoamine Oxidase A Gene (MAOA) Predicts melanogaster.” PLoS Genetics 2(9):1386–95. Behavioral Aggression Following Provocation.” Proceedings of Eguıluz,´ V. M., D. R. Chialvo, G. A. Cecchi, M. Baliki, and A. V. Ap- the National Academy of Sciences USA 106 (7): 2118–23. karian. 2005. “Scale-Free Brain Functional Networks.” Physical McNulty, J. E., C. M. Dowling, and M. H. Ariotti. 2009. “Driving Review Letters 94 (1): 1–4. Saints to Sin: How Increasing the Difficulty of Voting Dissuades Fagiolini, M., C. L. Jensen, and F. A. Champagne. 2009. “Epigenetic Even the Most Motivated Voters.” Political Analysis 17 (4): 435– Influences on Brain Development and Plasticity.” Current Opinion 55. in Neurobiology 19 (2): 207–12. Muotri, A. R., M. C. N. Marchetto, N. G. Coufal, R. Oefner, G. Yeo, Faulkner, G. J. 2011. “Retrotransposons: Mobile and Mutagenic from K. Nakashima, and F. H. Gage. 2010. “L1 Retrotransposition in Conception to Death.” FEBS Letters 585 (11): 1589–94. Neurons Is Modulated by MeCP2.” Nature 468 (7322): 443–46. Fontdevila, A. 2011. The Dynamic Genome: A Darwinian Approach. Mychasiuk, R., S. Zahir, N. Schmold, S. Ilnytskyy, O. Kovalchuk, and New York: Oxford University Press. R. Gibb. 2012. “Parental Enrichment and Offspring Development: Fowler, J. H., and C. T. Dawes. 2008. “Two Genes Predict Voter Modifications to Brain, Behavior and the Epigenome.” Behavioral Turnout.” Journal of Politics 70 (03): 579–94. and Brain Research 228 (2): 294–98. Frisco, M. L., C. Muller, and K. Dodson. 2004. “Participation in Nie, N. H., J. Junn, and K. Stehlik Barry. 1996. Education and Demo- Voluntary Youth-serving Associations and Early Adult Voting cratic Citizenship in America. Chicago: University of Chicago Behavior.” Social Science Quarterly 85 (3): 660–76. Press. “From Gene to Protein: Gene Expression Quantification Offers Noskova, T., N. Pivac, G. Nedic, A. Kazantseva, and D. M. Seler. New Insights.” 2011. PHYSorg.com. http://phys.org/news/2011-05- 2008. “Ethnic Differences in the Serotonin Transporter Polymor- gene-protein-quantification- insights.html. phism (5-HTTLPR) in Several European Populations.” Progress Garland, T., Jr, and S. A. Kelly. 2006. “Phenotypic Plasticity and in Neuro-Psychopharmacology: Biological Psychiatry 32 (7): 1735– Experimental Evolution.” Journal of Experimental Biology 209 39. (12): 2344–61. Notini, A. J., J. M. Craig, and S. J. White. 2008. “Copy Number Gelernter, J., D. Goldman, and N. Risch. 1993. “The A1 Allele at the Variation and Mosaicism.” Cytogenetic and Genome Research 123 D2 Dopamine Receptor Gene and Alcoholism: A Reappraisal.” (1–4): 270–77. Journal of the American Medical Association 269 (13): 1673–67. Pigliucci, M., C. J. Murren, and C. D. Schlichting. 2006. “Phenotypic Hao, K., E. Chudin, D. Greenawalt, and E. E. Schadt. 2010. “Mag- Plasticity and Evolution by Genetic Assimilation.” Journal of Ex- nitude of Stratification in Human Populations and Impacts on perimental Biology 209 (12): 2362–67. Genome Wide Association Studies.” PLoS ONE 5(1):e8695. Plenz, D. 2012. “Neuronal Avalanches and Coherence Potentials.” Hart, D., T. M. Donnelly, J. Youniss, and R. Atkins. 2007. “High European Physical Journal Special Topics 205 (1): 259–301. School Community Service as a Predictor of Adult Voting and Popper, K. R. 2002. The Logic of Scientific Discovery.London:Rout- Volunteering.” American Educational Research Journal 44 (1): ledge. 197–219. Price, A. L., A. Helgason, S. Palsson, and K. Stefansson. 2009. “The He, B. J., J. M. Zempel, A. Z. Snyder, and M. E. Raichle. 2010. Impact of Divergence Time on the Nature of Population Structure: “The Temporal Structures and Functional Significance of Scale- An Example from Iceland.” PLoS Genetics 5(6):e1000505. free Brain Activity.” Neuron 66 (3): 353–69. Price, A. L., N. A. Zaitlen, D. Reich, and N. Patterson. 2010. “New Helgason, A., B. Yngvadottir, B. Hrafnkelsson, J. Gulcher, and K. Approaches to Population Stratification in Genome-wide Associ- Stefansson. 2005. “An Icelandic Example of the Impact of Popu- ation Studies.” Nature Reviews Genetics 11 (7): 459–63. lation Structure on Association Studies.” Nature Genetics 37 (1): Proekt, A., J. R. Banavar, A. Maritan, and D. W. Pfaff. 2012. “Scale 90–95. Invariance in the Dynamics of Spontaneous Behavior.” Proceed- Iourov, I. Y., S. G. Vorsanova, and Y. B. Yurov. 2006. “Chromosomal ings of the National Academy of Sciences USA 109 (26): 10564–99. Variation in Mammalian Neuronal Cells: Known Facts and Attrac- Qureshi, I. A., and M. F. Mehler. 2012. “Emerging Roles of Non- tive Hypotheses.” International Review of Cytology 249: 143–91. coding RNAs in Brain Evolution, Development, Plasticity and Iskow, R. C., M. T. McCabe, R. E. Mills, S. Torene, W. S. Pittard, Disease.” Nature Reviews Neuroscience 13 (8): 528–41. A. F. Neuwald, E. G. Van Meir, P. M. Vertino, and S. E. Devine. Redon, R., S. Ishikawa, K. R. Fitch, L. Feuk, G. H. Perry, and T. D. 2010. “Natural Mutagenesis of Human Genomes by Endogenous Andrews. 2006. “Global Variation in Copy Number in the Human Retrotransposons.” Cell 141 (7): 1253–61. Genome.” Nature 444: 444–54. Julio, J. R. 2009. “Attitude-dependent Altruism, Turnout and Vot- Rehen, S. K. 2005. “Constitutional Aneuploidy in the Normal Human ing.” Public Choice 140 (1–2): 223–44. Brain.” Journal of Neuroscience 25: 2176–80. Kaushal, D., J. J. A. Contos, K. Treuner,A. H. Yang, M. A. Kingsbury, Schwanhausser,¨ B., D. Busse, N. Li, G. Dittmar, J. Schuchhardt, J. S. K. Rehen, M. J. McConnell, et al. 2003. “Alteration of Gene Wolf, W. Chen, and M. Selbach. 2011. “Global Quantification of Expression by Chromosome Loss in the Postnatal Mouse Brain.” Mammalian Gene Expression Control.” Nature 473 (7347): 337– Journal of Neuroscience 23 (13): 5599–606. 42.

13 Genopolitics and the Science of Genetics May 2013

Sciamanna, I., P. Vitullo, A. Curatolo, and C. Spadafora. 2009. van den Oord, E. J. C. G., F. C. Verhulst, and D. I. Boomsma. 1996. “Retrotransposons, Reverse Transcriptaseand the Genesis of New “A Genetic Study of Maternal and Paternal Ratings of Problem Genetic Information.” Gene 448 (2): 180–86. Behaviors in 3-year-old Twins.” Journal of Abnormal Psychology Shew, W. L., and D. Plenz. 2012. “The Functional Benefits of Criti- 105 (3): 349–57. cality in the Cortex.” Neuroscientist 19 (1): 88–100. Vitullo, P., I. Sciamanna, M. Baiocchi, P. Sinibaldi-Vallebona, Singer, T., M. J. McConnell, M. C. N. Marchetto, N. G. Coufal, and and C. Spadafora. 2012. “LINE-1 Retrotransposon Copies F. H. Gage. 2010. “LINE-1 Retrotransposons: Mediators of So- Are Amplified during Murine Early Embryo Develop- matic Variation in Neuronal Genomes?” Trends in Neurosciences ment.” Molecular Reproduction and Development 79 (2): 118– 33 (8): 345–54. 27. Skinner, Michael K. 2011. “Role of Epigenetics in Developmental Bi- West-Eberhard, M. J. 2003. Developmental Plasticity and Evolution. ology and Transgenerational Inheritance.” Birth Defects Research New York: Oxford University Press. Part C: Embryo Today: Reviews 93 (1): 51–55. “Study Suggests Westra, J. W., R. R. Rivera, D. M. Bushman, Y. C. Yung, S. E. Pe- Link between Oral Health and Ethnicity.” 2007. British Dentistry terson, S. Barral, and J. Chun. 2010. “Neuronal DNA Content Journal 202 (9): 514–15. Variation (DCV) with Regional and Individual Differences in the Theodore, W. H., S. S. Spencer, S. Wiebe, J. T. Langfitt, A. Ali, P. O. Human Brain.” Journal of Comparative Neurology 518 (19): 3981– Shafer, A. T. Berg, and B. G. Vickrey. 2006. “Epilepsy in North 4000. America: A Report Prepared under the Auspices of the Global Yashin, A. 2011. “Genetics of Exceptional Lifespan: Testing Muta- Campaign against Epilepsy, the International Bureau for Epilepsy, tion Accumulation Hypothesis of Aging Using GWAS” Rresented the International League against Epilepsy, and the World Health at Integrating Genetics and the Social Sciences conference, Boul- Organization.” Epilepsia 47 (10): 1700–22. der, CO. Thomas, D. C., and J. S. Witte. 2002. “Point: Population Stratification: Zhang, T. Y., and M. J. Meaney. 2010. “Epigenetics and AProblemforCase-ControlStudiesofCandidate-GeneAssoci- the Environmental Regulation of the Genome and Its ations?” Cancer Epidemiology, Biomarkers & Prevention 11 (6): Function.” Annual Review of Psychology 61 (1): 439– 505–12. 66. Tom, W. S. 1998. “A Review of Church Attendance Measures.” Zwarts, L., M. M. Magwire, M. A. Carbone, M. Versteven, L. American Sociological Review 63 (1): 131–36. Herteleer, R. R. H. Anholt, P. Callaerts, and T. F. C. Mackay. U.S. Census Bureau. 2012. “Voting-Age Population, Percent Re- 2011. “Complex Genetic Architecture of Drosophila Aggressive porting Registered, and Voted. 2012.” http://www.census.gov/ Behavior.” Proceedings of the National Academy of Sciences USA compendia/statab/2012/tables/12s0399.pdf. 108 (41): 17070–75.